Tracking transients in steelpan strikes using surveillance technology.

Area of Science:

• Acoustics and Musical Instrument Analysis
• Computer Vision and Image Processing
• Vibrational Spectroscopy

Background:

• Accurate measurement of vibrational modes in musical instruments is crucial for understanding their acoustic properties.
• Traditional methods for analyzing steelpan vibrations can be labor-intensive and may lack precision.
• Electronic speckle pattern interferometry (ESPI) offers a non-contact method for visualizing surface deformations.

Purpose of the Study:

• To develop and validate advanced computer vision techniques for precise feature tracking in high-speed steelpan videos.
• To improve the accuracy of interference fringe counting in electronic speckle pattern interferometry (ESPI) applied to musical instruments.
• To create a versatile model capable of analyzing vibrational patterns on various musical instrument surfaces.

Main Methods:

• Utilized robust computer vision libraries for object detection and image segmentation of steelpan surfaces in high-speed videos.
• Implemented data cleaning techniques for the training dataset to enhance model performance.
• Developed a segmentation-regression map for comprehensive drum surface analysis, complementing traditional object detection.

Main Results:

• Achieved a 10% or greater increase in the accuracy of fringe counts compared to previous methodologies.
• The segmentation-regression map provided interference fringe counts comparable to those obtained via object detection.
• Demonstrated the model's ability to generalize, counting fringes for instruments not included in the initial training set.

Conclusions:

• Advanced computer vision significantly enhances the accuracy of vibrational analysis in steelpans using ESPI.
• The developed segmentation-regression approach offers a robust and adaptable method for quantifying interference fringes on musical instruments.
• This work provides a foundation for more detailed acoustic studies of various musical instruments through improved image analysis.